Towards Precision Agriculture: A Unified CNN and Random Forest Framework for Jasmine Leaf Disease Recognition

The model's abstract performance characteristics of categorization across several classes are highlighted, with special attention paid to accuracy, F1-Score, recall, encouragement, and support percentage. The model is tested using ten distinct classes, known to be Significant Levels 1 through 10. Remember that accuracy and the F1-Score were recognized parameters for determining if a model was suitable for classification. Out of all the affirmative predictions, the accuracy percentages for each category vary between 89.87% to 93.49%, showing that the model can accurately recognize instances of each class. Recall values also span from 89.65% to 92.89%, indicating the percentage of true positive occurrences that the model correctly recognizes. The F1-Score, or harmonic average of accuracy and recall, indicates an acceptable assessment of the model's overall accuracy, ranging from 90.96% to 92.89%. Every class's frequency of occurrences is shown in the support column, which has values between 785 and 915. In the support % column, you can see the percentage of every class throughout the whole dataset. Overall model precision is represented by the percentage of successfully recognized occurrences (81.83%) across all classes. These three figures—Macro, Weighted, or Micro Average—indicate the overall effectiveness of the model. Macro Average yields 91.77% for the unweighted median of precision, recall, along with F1-Score. The weighted average produces a standard deviation of 91.79% by weighing the contribution of each class based on its support. At the score of 91.79%, the Micro Average displays the average accuracy over all instances. Across multiple classes, the model shows good accuracy, recall, or F1- Score; weighted averages highlight the balanced contribution of each class to overall performance. The model's claimed accuracy of 81.83% suggests that it can correctly classify cases over the whole dataset.